1 files changed, 25 insertions, 100 deletions

diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index a5e93b185515..7bdefd0c789a 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -645,9 +645,7 @@ void ceph_add_cap(struct inode *inode,
 	dout("add_cap %p mds%d cap %llx %s seq %d\n", inode,
 	     session->s_mds, cap_id, ceph_cap_string(issued), seq);
 
-	spin_lock(&session->s_gen_ttl_lock);
-	gen = session->s_cap_gen;
-	spin_unlock(&session->s_gen_ttl_lock);
+	gen = atomic_read(&session->s_cap_gen);
 
 	cap = __get_cap_for_mds(ci, mds);
 	if (!cap) {
@@ -785,10 +783,8 @@ static int __cap_is_valid(struct ceph_cap *cap)
 	unsigned long ttl;
 	u32 gen;
 
-	spin_lock(&cap->session->s_gen_ttl_lock);
-	gen = cap->session->s_cap_gen;
+	gen = atomic_read(&cap->session->s_cap_gen);
 	ttl = cap->session->s_cap_ttl;
-	spin_unlock(&cap->session->s_gen_ttl_lock);
 
 	if (cap->cap_gen < gen || time_after_eq(jiffies, ttl)) {
 		dout("__cap_is_valid %p cap %p issued %s "
@@ -1182,7 +1178,8 @@ void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
 	 * s_cap_gen while session is in the reconnect state.
 	 */
 	if (queue_release &&
-	    (!session->s_cap_reconnect || cap->cap_gen == session->s_cap_gen)) {
+	    (!session->s_cap_reconnect ||
+	     cap->cap_gen == atomic_read(&session->s_cap_gen))) {
 		cap->queue_release = 1;
 		if (removed) {
 			__ceph_queue_cap_release(session, cap);
@@ -1534,7 +1531,7 @@ static inline int __send_flush_snap(struct inode *inode,
  * asynchronously back to the MDS once sync writes complete and dirty
  * data is written out.
  *
- * Called under i_ceph_lock.  Takes s_mutex as needed.
+ * Called under i_ceph_lock.
  */
 static void __ceph_flush_snaps(struct ceph_inode_info *ci,
 			       struct ceph_mds_session *session)
@@ -1656,7 +1653,6 @@ retry:
 	mds = ci->i_auth_cap->session->s_mds;
 	if (session && session->s_mds != mds) {
 		dout(" oops, wrong session %p mutex\n", session);
-		mutex_unlock(&session->s_mutex);
 		ceph_put_mds_session(session);
 		session = NULL;
 	}
@@ -1665,10 +1661,6 @@ retry:
 		mutex_lock(&mdsc->mutex);
 		session = __ceph_lookup_mds_session(mdsc, mds);
 		mutex_unlock(&mdsc->mutex);
-		if (session) {
-			dout(" inverting session/ino locks on %p\n", session);
-			mutex_lock(&session->s_mutex);
-		}
 		goto retry;
 	}
 
@@ -1680,12 +1672,10 @@ retry:
 out:
 	spin_unlock(&ci->i_ceph_lock);
 
-	if (psession) {
+	if (psession)
 		*psession = session;
-	} else if (session) {
-		mutex_unlock(&session->s_mutex);
+	else
 		ceph_put_mds_session(session);
-	}
 	/* we flushed them all; remove this inode from the queue */
 	spin_lock(&mdsc->snap_flush_lock);
 	list_del_init(&ci->i_snap_flush_item);
@@ -1915,7 +1905,6 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 	struct ceph_cap *cap;
 	u64 flush_tid, oldest_flush_tid;
 	int file_wanted, used, cap_used;
-	int took_snap_rwsem = 0;             /* true if mdsc->snap_rwsem held */
 	int issued, implemented, want, retain, revoking, flushing = 0;
 	int mds = -1;   /* keep track of how far we've gone through i_caps list
 			   to avoid an infinite loop on retry */
@@ -1923,14 +1912,13 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags,
 	bool queue_invalidate = false;
 	bool tried_invalidate = false;
 
+	if (session)
+		ceph_get_mds_session(session);
+
 	spin_lock(&ci->i_ceph_lock);
 	if (ci->i_ceph_flags & CEPH_I_FLUSH)
 		flags |= CHECK_CAPS_FLUSH;
-
-	goto retry_locked;
 retry:
-	spin_lock(&ci->i_ceph_lock);
-retry_locked:
 	/* Caps wanted by virtue of active open files. */
 	file_wanted = __ceph_caps_file_wanted(ci);
 
@@ -2010,7 +1998,7 @@ retry_locked:
 			ci->i_rdcache_revoking = ci->i_rdcache_gen;
 		}
 		tried_invalidate = true;
-		goto retry_locked;
+		goto retry;
 	}
 
 	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
@@ -2024,8 +2012,6 @@ retry_locked:
 		    ((flags & CHECK_CAPS_AUTHONLY) && cap != ci->i_auth_cap))
 			continue;
 
-		/* NOTE: no side-effects allowed, until we take s_mutex */
-
 		/*
 		 * If we have an auth cap, we don't need to consider any
 		 * overlapping caps as used.
@@ -2088,37 +2074,8 @@ retry_locked:
 			continue;     /* nope, all good */
 
 ack:
-		if (session && session != cap->session) {
-			dout("oops, wrong session %p mutex\n", session);
-			mutex_unlock(&session->s_mutex);
-			session = NULL;
-		}
-		if (!session) {
-			session = cap->session;
-			if (mutex_trylock(&session->s_mutex) == 0) {
-				dout("inverting session/ino locks on %p\n",
-				     session);
-				session = ceph_get_mds_session(session);
-				spin_unlock(&ci->i_ceph_lock);
-				if (took_snap_rwsem) {
-					up_read(&mdsc->snap_rwsem);
-					took_snap_rwsem = 0;
-				}
-				if (session) {
-					mutex_lock(&session->s_mutex);
-					ceph_put_mds_session(session);
-				} else {
-					/*
-					 * Because we take the reference while
-					 * holding the i_ceph_lock, it should
-					 * never be NULL. Throw a warning if it
-					 * ever is.
-					 */
-					WARN_ON_ONCE(true);
-				}
-				goto retry;
-			}
-		}
+		ceph_put_mds_session(session);
+		session = ceph_get_mds_session(cap->session);
 
 		/* kick flushing and flush snaps before sending normal
 		 * cap message */
@@ -2130,20 +2087,7 @@ ack:
 			if (ci->i_ceph_flags & CEPH_I_FLUSH_SNAPS)
 				__ceph_flush_snaps(ci, session);
 
-			goto retry_locked;
-		}
-
-		/* take snap_rwsem after session mutex */
-		if (!took_snap_rwsem) {
-			if (down_read_trylock(&mdsc->snap_rwsem) == 0) {
-				dout("inverting snap/in locks on %p\n",
-				     inode);
-				spin_unlock(&ci->i_ceph_lock);
-				down_read(&mdsc->snap_rwsem);
-				took_snap_rwsem = 1;
-				goto retry;
-			}
-			took_snap_rwsem = 1;
+			goto retry;
 		}
 
 		if (cap == ci->i_auth_cap && ci->i_dirty_caps) {
@@ -2165,9 +2109,10 @@ ack:
 
 		__prep_cap(&arg, cap, CEPH_CAP_OP_UPDATE, mflags, cap_used,
 			   want, retain, flushing, flush_tid, oldest_flush_tid);
-		spin_unlock(&ci->i_ceph_lock);
 
+		spin_unlock(&ci->i_ceph_lock);
 		__send_cap(&arg, ci);
+		spin_lock(&ci->i_ceph_lock);
 
 		goto retry; /* retake i_ceph_lock and restart our cap scan. */
 	}
@@ -2182,13 +2127,9 @@ ack:
 
 	spin_unlock(&ci->i_ceph_lock);
 
+	ceph_put_mds_session(session);
 	if (queue_invalidate)
 		ceph_queue_invalidate(inode);
-
-	if (session)
-		mutex_unlock(&session->s_mutex);
-	if (took_snap_rwsem)
-		up_read(&mdsc->snap_rwsem);
 }
 
 /*
@@ -2198,26 +2139,17 @@ static int try_flush_caps(struct inode *inode, u64 *ptid)
 {
 	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
 	struct ceph_inode_info *ci = ceph_inode(inode);
-	struct ceph_mds_session *session = NULL;
 	int flushing = 0;
 	u64 flush_tid = 0, oldest_flush_tid = 0;
 
-retry:
 	spin_lock(&ci->i_ceph_lock);
 retry_locked:
 	if (ci->i_dirty_caps && ci->i_auth_cap) {
 		struct ceph_cap *cap = ci->i_auth_cap;
 		struct cap_msg_args arg;
+		struct ceph_mds_session *session = cap->session;
 
-		if (session != cap->session) {
-			spin_unlock(&ci->i_ceph_lock);
-			if (session)
-				mutex_unlock(&session->s_mutex);
-			session = cap->session;
-			mutex_lock(&session->s_mutex);
-			goto retry;
-		}
-		if (cap->session->s_state < CEPH_MDS_SESSION_OPEN) {
+		if (session->s_state < CEPH_MDS_SESSION_OPEN) {
 			spin_unlock(&ci->i_ceph_lock);
 			goto out;
 		}
@@ -2254,9 +2186,6 @@ retry_locked:
 		spin_unlock(&ci->i_ceph_lock);
 	}
 out:
-	if (session)
-		mutex_unlock(&session->s_mutex);
-
 	*ptid = flush_tid;
 	return flushing;
 }
@@ -3213,8 +3142,7 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 	if (complete_capsnap)
 		wake_up_all(&ci->i_cap_wq);
 	while (put-- > 0) {
-		/* avoid calling iput_final() in osd dispatch threads */
-		ceph_async_iput(inode);
+		iput(inode);
 	}
 }
 
@@ -3288,7 +3216,7 @@ static void handle_cap_grant(struct inode *inode,
 	u64 size = le64_to_cpu(grant->size);
 	u64 max_size = le64_to_cpu(grant->max_size);
 	unsigned char check_caps = 0;
-	bool was_stale = cap->cap_gen < session->s_cap_gen;
+	bool was_stale = cap->cap_gen < atomic_read(&session->s_cap_gen);
 	bool wake = false;
 	bool writeback = false;
 	bool queue_trunc = false;
@@ -3340,7 +3268,7 @@ static void handle_cap_grant(struct inode *inode,
 	}
 
 	/* side effects now are allowed */
-	cap->cap_gen = session->s_cap_gen;
+	cap->cap_gen = atomic_read(&session->s_cap_gen);
 	cap->seq = seq;
 
 	__check_cap_issue(ci, cap, newcaps);
@@ -3553,13 +3481,12 @@ static void handle_cap_grant(struct inode *inode,
 	if (wake)
 		wake_up_all(&ci->i_cap_wq);
 
+	mutex_unlock(&session->s_mutex);
 	if (check_caps == 1)
 		ceph_check_caps(ci, CHECK_CAPS_AUTHONLY | CHECK_CAPS_NOINVAL,
 				session);
 	else if (check_caps == 2)
 		ceph_check_caps(ci, CHECK_CAPS_NOINVAL, session);
-	else
-		mutex_unlock(&session->s_mutex);
 }
 
 /*
@@ -4203,8 +4130,7 @@ done:
 	mutex_unlock(&session->s_mutex);
 done_unlocked:
 	ceph_put_string(extra_info.pool_ns);
-	/* avoid calling iput_final() in mds dispatch threads */
-	ceph_async_iput(inode);
+	iput(inode);
 	return;
 
 flush_cap_releases:
@@ -4246,8 +4172,7 @@ void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
 			spin_unlock(&mdsc->cap_delay_lock);
 			dout("check_delayed_caps on %p\n", inode);
 			ceph_check_caps(ci, 0, NULL);
-			/* avoid calling iput_final() in tick thread */
-			ceph_async_iput(inode);
+			iput(inode);
 			spin_lock(&mdsc->cap_delay_lock);
 		}
 	}